INTRODUCTION

Kaposi's sarcoma (KS) is a vascular proliferative disorder with four epidemiological forms: classic, African endemic, iatrogenic transplant related, and epidemic acute immune deficiency syndrome (AIDS) related (1). KS is usually limited to the skin but may involve mucus membranes, visceral organs, and lymph nodes in the more aggressive forms (2). For cutaneous KS, a histopathological progression from patch to plaque and nodular stages correlates with a clinical progression from macules to palpable tumors (2). The patch stage is characterized by networks of thin-walled irregular endothelial-lined spaces; the plaque stage, by sparse spindle cells intermingled among irregular endothelial-lined spaces; and the nodular stage, by sheets or fascicles of spindle cells (2).

Although the exact pathogenesis of KS is not defined, it has been suggested that a chronic cytokine-mediated proliferative state of an endothelial cell, possibly triggered by activation of a latent KS-associated herpes virus (KSHV) infection, promotes transformation to a neoplastic cell (1, 3, 4). The origin of KS from blood vessel or lymphatic endothelium is controversial. Although some early studies failed to detect the presence of blood vessel endothelial markers in KS (5), other studies reported the presence of traditional blood vessel endothelial markers including CD31, CD34, and von Willebrand factor (vWF, Factor VIII–related antigen) (6, 7, 8, 9). The latter findings are compatible with a theory of origin of KS from blood vessel endothelium. However, this interpretation is complicated by recent observations that the above blood vessel endothelial markers are also weakly expressed on lymphatic endothelium (9, 10, 11). Furthermore, ultrastructural features and the absence of the blood vessel endothelium–specific PAL-E antigen (12) from KS argue against an origin from blood vessel endothelial cells (13, 14, 15). These considerations have led some authors to suggest that KS originates from lymphatic endothelium (5, 14). This notion has been reinforced recently by the demonstration of a uniform expression of two lymphatic endothelial markers, vascular endothelial growth factor receptor-3 (VEGFR-3; 16) and podoplanin (17), in KS (18, 19, 20). However, VEGFR-3 has also been shown to be expressed by neoplastic blood vessel endothelia in both benign and malignant vascular tumors and, in this situation, loses its selectivity for lymphatic endothelium (21). Therefore, we wished to reexamine these results with an independent marker of lymphatic endothelium.

We recently observed that D2-40, a monoclonal antibody (mAb) to an Mr 40,000 O-linked sialoglycoprotein (22), was a selective marker of lymphatic endothelium. In this study, we examined the expression of this novel marker of lymphatic endothelium in normal tissues and compared its reactivity with the expression of CD31 in a series of 62 vascular lesions, including 24 cases of cutaneous KS at different stages of progression.

MATERIALS AND METHODS

Formalin-fixed, paraffin-embedded blocks of normal tissue, 62 vascular lesions (24 cutaneous KS, 7 angiosarcomas, 10 lymphangiomas, 10 hemangiomas, 3 glomus tumors, 2 angiolipomas, 2 pyogenic granulomas, 2 vascular malformations, 1 hemangiopericytoma, 1 hemangioendothelioma), 28 nonvascular soft-tissue tumors (10 leiomyosarcomas, 6 neurofibromas, 6 dermatofibromas, 4 dermatofibrosarcoma protuberans, 2 malignant fibrous histiocytomas), as well as frozen samples of tonsil, were obtained from the Department of Pathology, Sunnybrook and Women's College Health Sciences Center, University of Toronto, Ontario, Canada. The KS specimens included patch, plaque, and nodular lesions from patients with classic and AIDS- and transplant-related KS. None of the angiosarcomas were associated with lymphedema or preexisting lymphangioma.

MAb D2-40 (IgG1) was purified from mouse ascitic fluid, as previously described (22), and stored at a concentration of 0.7 mg/mL under sterile conditions at 4° C. The mAb to an undefined blood vessel endothelial antigen, clone PAL-E, (hybridoma supernatant 15 μg/mL from Monosan), whose reactivity is restricted to frozen sections of tissues, was purchased through Cederlane Laboratories (Hornby, Ontario, Canada). The mAb to CD31 (Clone JC/70A, hybridoma supernatant) was purchased from DAKO (Carpinteria, CA).

Sections of paraffin-embedded tissues (5 μm) were dewaxed through graded concentrations of ethanol. For immunostaining, heat-inducted epitope retrieval by microwave pretreatment (800 W for 10 min) was only performed for CD31. The sections were first incubated in methanol containing 3% H2O2 to inactivate endogenous peroxidase. The sections were then incubated with mAb D2-40 (0.1 μg/mL), or anti-CD31 (1:40 dilution), followed sequentially by biotinylated goat anti-mouse immunoglobulin (IgG) antibody (Zymed, San Francisco, CA) at a 1:200 dilution and a horseradish peroxidase–avidin conjugate (DAKO) at a 1:500 dilution. For color development, the sections were incubated with 3,3′-diamino-benzidine. Adjacent cryostat sections (5 μm) of quick-frozen tissues were air-dried and fixed in cold acetone for 10 min. After inactivation of the endogenous peroxidase as described above, the sections were immunostained with mAb PAL-E (0.15 μg/mL) or D2-40 (0.1 μg/mL), also as described above.

RESULTS

Consecutive cryostat sections of human tonsil were stained with mAb D2-40 or PAL-E (that is specific for endothelial cells of blood vessels; 12). These antibodies stained nonoverlapping populations of thin-walled endothelial channels in adjacent sections. PAL-E stained blood vessels (Fig. 1A, arrowheads) that did not stain with D2-40 (Fig. 1B). Conversely, D2-40 stained lymphatic channels (Fig. 1B, arrowheads) that did not stain with PAL-E (Fig. 1A). In paraffin sections of normal breast, D2-40 stained the endothelium of lymphatic channels (Fig. 1C, arrowheads), but not the endothelium of an artery in a vascular bundle (Fig. 1C, arrow).

FIGURE 1
figure 1

Selective staining of lymphatic vessels with D2-40 in normal tissues. A and B, consecutive frozen sections of tonsil were stained with PAL-E (A) or D2-40 (B), showing positive endothelial staining of distinct populations of vessels (arrowheads). C and D, paraffin sections of breast were stained with D2-40 showing positive staining of lymphatics (arrowheads) and absence of staining of artery (C) or capillary (D; arrows). Note the presence of lymphocytes in the lumen of the lymphatic vessel in (D). (Original magnification, A and B: 100 ×; C: 300 ×; D: 400 ×).

Full size image

D2-40 also stained the endothelium of a lymphatic channel containing lymphocytes (Fig. 1D, arrowhead), but not the adjacent capillary (Fig. 1D, arrow). The results of immunostaining of a series of vascular lesions with mAb D2-40 and for CD31 are summarized in Table 1. D2-40 reacted strongly with all cases of cutaneous KS (24/24) at the patch (Fig. 2B–C), plaque, and nodular (Fig. 2E–F) stages of progression. D2-40 also reacted with the endothelium lining the sinusoidal spaces in all lymphangiomas (10/10; Fig. 3A) but did not react with any hemangiomas (0/10), glomus tumors (0/3), angiolipomas (0/2), pyogenic granulomas (0/2), vascular malformations (0/2), hemangiopericytoma (0/1), or hemangioendothelioma (0/1). Among angiosarcomas, there was diffuse immunostaining of the neoplastic endothelial cells in 3/7 cases (Fig. 3C) and absence of staining in 4/7 cases (Fig. 3B). Of the three D2-40-positive angisarcomas, all exhibited hobnail endothelial cells, and one had, in addition, interspersed lymphoid aggregates. None of these features were seen in D2-40-negative angiosarcomas. With respect to nonvascular soft-tissue tumors, D2-40 did not stain neurofibromas (0/6) or dermatofibrosarcoma protuberans (0/4) and reacted with a fraction of leiomyosarcomas (3/10), dermatofibromas (2/6), and malignant fibrous histiocytomas (1/2). CD31 was positive in all cases of KS, hemangioma, angiolipoma, pyogenic granuloma, vascular malformation, hemangiopericytoma, hemangioendothelioma, and angiosarcoma (Fig. 3E–F). CD31 did not stain glomus tumors (0/3). Among lymphangiomas, there was absence of staining for CD31 in 5/10 cases (Fig. 3D) and weak staining in 5/10 cases.

TABLE 1 Expression of D2-40 and CD31 in 62 Vascular Lesions
Full size table
FIGURE 2
figure 2

Staining of Kaposi's sarcoma with D2-40. A–C, patch-stage Kaposi's sarcoma stained with hematoxylin and eosin (A) or with D2-40 (B and C). There is diffuse positive staining of the irregular endothelial lined spaces with D2-40 in B and C. D–F, nodular-stage Kaposi's sarcoma stained with hematoxylin and eosin (D) or D2-40 (E and F). There is diffuse positive staining of the spindle cells with D2-40 in E and F (original magnification, A, B, D, and E: 100 ×; C and F, 300 ×).

Full size image
FIGURE 3
figure 3

Staining of lymphangioma and angiosarcoma with D2-40 or anti-CD31. AandD, consecutive paraffin sections of a lymphangioma were stained with D2-40 (A), showing strong positive staining of sinusoidal spaces, or with anti-CD31 (D), showing absence of staining. BandE, consecutive paraffin sections of an angiosarcoma were stained with D2-40 (B), showing absence of staining, or with anti-CD31 (E), showing diffuse positive staining. Cand F, consecutive paraffin sections of an angiosarcoma were stained with D2-40 (C), showing diffuse positive staining, or with anti-CD31 (F), likewise showing diffuse positive staining. Note absence of staining of the endothelial lining of a normal artery with D2-40 in C (arrow) and positive staining with anti-CD31 in F (arrowhead; original magnification, A to F: 100 ×).

Full size image

DISCUSSION

D2-40 was previously reported to react with an oncofetal antigen expressed in fetal testis and on the surface of testicular germ cell tumors, but not in adult testis (22). In this study, we report, for the first time, that this mAb is also a new selective marker of lymphatic endothelium and does not react with blood vessel endothelium, as demonstrated by the following results.

First, D2-40 reacted with the endothelium of lymphatic channels presumptively characterized by their morphological appearance in normal tissues. Specifically, these were flattened or open spaces lined by a single layer of endothelial cells whose lumen was sometimes filled with lymphocytes. The endothelium of blood vessels including arteries, veins, or capillaries did not stain with D2-40. To definitively distinguish thin-walled lymphatics from capillaries and venules, we compared the immunostaining of adjacent frozen sections of tonsil with mAb D2-40 and PAL-E, a recognized specific marker of blood vessel endothelium whose reactivity is limited to frozen sections of tissues (12). In these sections, D2-40 and PAL-E stained nonoverlapping populations of vessels, confirming the selective reactivity of D2-40 with lymphatic channels.

D2-40 reacted effectively with a fixation-resistant epitope on lymphatic endothelial cells in conventionally processed formalin-fixed and paraffin-embedded tissues without the requirement for epitope retrieval. This facilitated our use of this reagent to survey a series of vascular tumors and tumor-like lesions from our archival collection. Initially, we demonstrated that the selectivity of D2-40 for lymphatic endothelium in normal tissues was maintained for the neoplastic and reactive endothelium of benign vascular lesions. In particular, D2-40 reacted with the endothelial cells lining the sinusoidal spaces of lymphangiomas, (10/10, 100% sensitivity), but not the endothelial cells of hemangiomas (0/10, 100% specificity) or other vascular lesions of blood vessel origin, including glomus tumors (0/3), angiolipomas (0/2), pyogenic granulomas (0/2), vascular malformations (0/2), hemangiopericytoma (0/1), or hemangioendothelioma (0/1, 100% specificity). The endothelial antigen, CD31, which is expressed uniformly on the endothelium of blood vessels but only focally on lymphatics (10, 11), was present in all vascular lesions examined, with the exception of glomus tumors (0/3) and lymphangiomas, where it was not detected on the endothelium lining the sinusoidal spaces in 50% (5/10) of the cases. After establishing the selectivity of D2-40 for lymphatic endothelium in both normal tissues and benign vascular tumors and tumor-like lesions, we surveyed a series of KS and angiosarcomas for reactivity with this antibody.

All cases of cutaneous KS at different stages of progression reacted strongly with D2-40 (24/24, 100% sensitivity). D2-40 stained 100% of the endothelial cells lining the irregular spaces in patch- and plaque-stage lesions and the spindle cells in the plaque- and nodular-stage tumors. All cases of KS were also weakly positive for CD31 expression. These results suggest that KS originates from lymphatic endothelium, or, at the very least, from a pluripotent cell type capable of differentiating along both the lymphatic and blood vessel endothelial lineages, in agreement with similar conclusions reached on the basis of expression of VEGFR-3 and podoplanin (18, 19, 20). In the case of VEGFR-3, the interpretation of the results is complicated by the fact that this marker is selective for lymphatic endothelium in normal tissues, but not in vascular tumors, where it reacts with the neoplastic endothelium of benign and malignant blood vesslels (21).

Interestingly, our results with angiosarcomas indicate that a substantial subset (3/7) of these malignant vascular tumors can initiate a differentiation program along the lymphatic endothelial lineage based on their immunoreactivity with D2-40, again in agreement with similar conclusions based on the expression of the lymphatic endothelial markers VEGFR-3, podoplanin, and the β-chemokine receptor D6 (17, 20, 23). In the sample of angiosarcomas surveyed in our study, there was a clear distinction between the tumors that showed absence of staining with D2-40 and those that showed diffuse staining (100% of tumor cells positive) with the antibody. All the angiosarcomas examined in our study stained strongly for CD31 expression. Only the D2-40–positive angiosarcomas exhibited hobnail endothelial cells, as previously reported for VEGFR-3– positive angiosarcomas (20). Our results of D2-40 and CD31 staining suggest that angiosarcomas can be divided into subsets that originate from one of two distinct progenitor cell types, one restricted to differentiate along the blood vessel endothelial lineage and the other capable of differentiating along both the lymphatic and blood vessel lineages. Thus, a panel of markers of lymphatic endothelium including D2-40 would be useful to distinguish these tumors.

Among vascular lesions that could be considered in the differential diagnosis of KS, D2-40 did not react with hemangiomas (0/10), vascular malformations (0/2), hemangiopericytoma (0/1), or hemangioendothelioma (0/1) and reacted with 3/7 angiosarcomas. Therefore, with the exception of those angiosarcomas that express D2-40, this antibody can be used practically to distinguish these lesions from KS. With respect to nonvascular soft-tissue tumors, D2-40 did not react with neurofibromas (0/6) or dermatofibrosarcoma protuberans (0/4) and reacted with a fraction of leiomyosarcomas (3/10), dermatofibromas (2/6), and malignant fibrous histiocytomas (1/2). This indicates that D2-40 is selective for lymphatic versus vascular endothelium in both normal tissues and vascular tumors and tumor-like lesions but is not absolutely specific for lymphatic endothelium because it also reacts with a subset of nonvascular soft-tissue tumors of mesenchymal origin.

In summary, we report that D2-40 is a new selective monoclonal marker of lymphatic endothelium and an excellent immunohistochemical marker of KS. D2-40 also stains a subset of angiosarcomas and can be used in a panel of markers to subtype these tumors on the basis of their capacity to initiate a program of differentiation along the lymphatic endothelial lineage.